Biogeochemistry (2018) 141:463–486

https://doi.org/10.1007/s10533-018-0514-2 (0123456789().,-volV)(0123456789().,-volV)

Episodic salinization and freshwater salinization syndrome mobilize base cations, carbon, and nutrients to streams across urban regions

Shahan Haq . Sujay S. Kaushal . Shuiwang Duan

Received: 26 January 2018 / Accepted: 11 October 2018 / Published online: 24 October 2018 Ó Springer Nature Switzerland AG 2018

Abstract Urbanized watersheds in colder climates statistically significant linear increasing trends in experience episodic salinization due to anthropogenic calcium and potassium concentrations with experi- salt inputs and runoff from impervious surfaces. mental salinization across all 12 sites and in magne- Episodic salinization can be manifested as a ‘pulse’ sium concentrations at 11 of 12 sites (p \ 0.05), with in concentrations and fluxes of salt ions lasting from mean rates of increase of 1.92 ± 0.31 mg-Ca per hours to days after snowstorms in response to road g-NaCl, 2.80 ± 0.67 mg–K per g-NaCl, and salting. Episodic salinization contributes to freshwater 1.11 ± 0.19 mg-Mg per g-NaCl, respectively. Simi- salinization syndrome, characterized by cascading larly, there were statistically significant increasing mobilization of chemicals and shifting acid–base linear trends in total dissolved nitrogen (TDN) con- status. We conducted laboratory experiments and centrations with experimental salinization at 9 of the analyzed high-frequency sensor data to investigate the 12 sites, with a mean rate of increase of water quality impacts of freshwater salinization syn- 0.07 ± 0.01 mg-N per g-NaCl. There were statisti- drome and episodic salinization across 12 watersheds cally significant increasing linear trends in soluble draining two major metropolitan regions along the reactive phosphorus (SRP) concentrations with exper- U.S. East Coast. Sediments from 12 watersheds imental salinization at 7 of the 12 sites (p \ 0.05), spanning land use gradients across two metropolitan with a mean rate of increase of 2.34 ± 0.66 lg-P per regions, Baltimore, Maryland and Washington DC, g-NaCl. The response of dissolved inorganic carbon were incubated across a range of replicated salinity (DIC) and organic carbon (DOC) concentrations to treatments (0–10 g/L sodium chloride). There were experimental salinization varied between sites, and dissolved silica did not show any significant response. High-frequency sensors near the experimental sites Responsible Editor: Susana Bernal. showed statistically significant positive linear rela- tionships between nitrate concentrations, specific Electronic supplementary material The online version of conductance, and chloride concentrations similar to this article (https://doi.org/10.1007/s10533-018-0514-2) con- relationships observed in laboratory incubations. Our tains supplementary material, which is available to authorized users. results suggested that episodic salinization and fresh- water salinization syndrome can mobilize base cations S. Haq (&) Á S. S. Kaushal Á S. Duan and nutrients to streams through accelerated ion Department of Geology & Earth System Science exchange and stimulate different biogeochemical Interdisciplinary Center, University of Maryland, College Park, MD 20742, USA processes by shifting pH ranges and ionic strength. e-mail: [email protected] 123 464 Biogeochemistry (2018) 141:463–486

The growing impacts of freshwater salinization syn- episodic salinization on mobilization of base cations, drome and episodic salinization on nutrient mobiliza- carbon, and nutrients from sediments to stream water tion, shifting acid–base status, and augmenting using a combination of experimental incubations and eutrophication warrant serious consideration in water observations in the field from high-frequency sensor quality management. data. Due to their ionic nature, most salts are retained in Keywords Salt pollution Á Emerging contaminants Á soils and groundwater (Cooper et al. 2014; Findlay and Human-accelerated weathering Kelly 2011). As such, salts accumulate in the water- shed, and long-term salinization has been reported in rivers on most continents and biomes (Williams 2001; Kaushal et al. 2014a, b; Herbert et al. 2015). Urban Introduction areas are particularly vulnerable to salinization due to the combination of salt inputs (e.g., road salts, Many streams and rivers in the US and elsewhere are gypsum), impervious surfaces, and drainage infras- experiencing increased salinization due to salt pollu- tructure (Snodgrass et al. 2017; Kaushal et al. 2017; tion from anthropogenic inputs and accelerated Marsalek 2003). In the Northeastern USA, for exam- weathering in human-impacted watersheds (e.g., ple, salinity has doubled from 1990 to 2011 and has Kaushal et al. 2005; Canedo-Arguelles et al. 2013; exceeded the rate of urbanization (Corsi et al. 2015). Corsi et al. 2015; Kaushal et al. 2017). This saliniza- Previous studies suggest that rural streams and rivers tion can manifest itself as chronically high concentra- in the Eastern USA with as little as 5 percent tions of a mixture of salts throughout all seasons or impervious surface coverage within their watersheds episodically high concentrations contributing to a can also be at significant risk for long-term salinization ‘freshwater salinization syndrome’ on a continental (Kelly et al. 2008; Conway 2007), while in the scale (Kaushal et al. 2018). The freshwater saliniza- Midwestern USA, recent studies suggest that lakes tion syndrome impacts biodiversity, contaminant with as little as one percent impervious surface mobility, built infrastructure, and drinking water coverage are at risk (Dugan et al. 2017). quality (e.g., Corsi et al. 2015; Ramakrishna and Episodic and long-term salinization of fresh water Viraraghavan 2005; Stets et al. 2018; Novotny et al. can have ecosystem scale impacts such as the loss of 1998). Natural sources of salinity include the chemical native vegetation, disruptions in food webs, and the weathering of rock and soils throughout watersheds, mobilization of contaminants (e.g., heavy metals) dissolved ions in precipitation, and sea spray aerosols (Lofgren 2001; Norrstrom 2005; Backstrom et al. in coastal areas (Meybeck 2003; Kaushal et al. 2013). 2004; Amrhein et al. 1992). However, less is known Anthropogenic activities can dramatically increase the regarding impacts of salinization on carbon and salinity of water both directly (e.g., inputs of road nutrient cycles in fresh water (Duan and Kaushal salts, mining waste, industrial detergents, fertilizer 2015). Episodic salinization can enhance the mobi- salts, sewage discharges) (Kaushal et al. 2018), and lization of carbon, nutrients, and cations due to indirectly (e.g. inducing acid rain, building concrete coupled biotic and abiotic processes, such as ion and limestone infrastructure, increasing flood fre- exchange, rapid nitrification, pH, increased ionic quency, and changing land-use) (Barnes and Raymond strength, organic matter dispersion, and chloride 2009; Kaushal et al. 2017; Steele and Aitkenhead- complexation. While billions of dollars have been Peterson 2011). Although less considered, dissolved spent to reduce nitrogen and phosphorus loading by salts influence acid neutralizing capacity, pH, and stream and riparian restoration (Bernhardt et al. 2005), nutrient mobilization in watersheds (Kaushal et al. salt pollution has been largely unmanaged and unreg- 2018; Green et al. 2008; Compton and Church 2011). ulated. Part of the reason why effects of salinization on This can occur via ion exchange or complexation carbon and nutrient dynamics in streams and rivers is reactions, organic matter dispersion, or by alterations not well understood may be the different temporal and in microbial processes (Duan and Kaushal 2015; Corsi spatial scales at which salinization can occur. For et al. 2010;Oren2001; Kim and Koretsky 2013). Here, example, salinity (e.g., chloride) concentrations in we explore the potential effects of urban rivers may temporarily reach up to 33% the 123 Biogeochemistry (2018) 141:463–486 465 salinity of sea water in the hours to days following a salinization from sediments to streams. Sediments, snow storm, and then gradually accumulate in water- streamwater, and sodium chloride were incubated in a shed sinks (e.g., groundwater, soil, lakes) (Kaushal controlled lab environment to mimic post snowstorm et al. 2005; Kelly et al. 2008; Cooper et al. 2014). conditions (i.e., high salinity runoff entering the Acute and episodic salinization occurs over the course stream). These methods for incubations were previ- of hours to days following a snow event (e.g., road salt ously described in Duan and Kaushal (2015). We also pulse) and chronic long-term salinization can occur explored high-frequency sensor data from the U.S. over the course of seasons to decades (Kaushal et al. Geological Survey in the same vicinity as our labo- 2018). Some water quality effects of salinization, such ratory experiments sites to investigate whether there as nutrient and cation mobilization, would likely occur were similar relationships between specific conduc- over the course of several hours and would likely tance (a proxy or surrogate for dissolved salts) and immediately follow increases in salinization (e.g., nitrate (a key pollutant in the Chesapeake Bay post-storm). However, these effects may not be watershed leading to eutrophication). captured by traditional water quality monitoring, which is conducted by sampling a fixed point in a Site description stream or river on a weekly to monthly interval (as better explained in Rode et al. 2016; Pellerin et al. Sediment and streamwater were incubated form 12 2012). As such, these effects may be explored with a sites in the Baltimore-Washington Metropolitan Area combination of lab experiments and high-frequency in the Chesapeake Bay Watershed. Eight of the stream sensor data. sites are within the U.S. National Science Foundation The goal of this study was to investigate the (NSF) supported Baltimore Ecosystem Study Long potential water quality effects of episodic salinization Term Ecological Research Project (BES-LTER), and and freshwater salinization syndrome in urban water- are long-term, routinely monitored, and well charac- sheds induced by acutely elevated salinity across the terized sites (e.g., Groffman et al. 2004; Shields et al. sediment–water interface. We compared results of 2008; Duan et al. 2012; Majcher et al. 2018). These 8 experimental salinization in the laboratory with pat- stream sites are located in Baltimore, Maryland, where terns of episodic salinization (specific conductance they exhibit a land use gradient (Fig. 1). These sites and chloride) and nutrient loading (nitrate) from high- vary in drainage area (8–8350 ha), percent of water- frequency field sensors. Sensor data were analyzed to shed area covered in impervious surfaces (0–61%), provide additional insights on the timescales of population density (0 people/ha to 20 people/ha), and salinity-nutrient interactions, underlying mechanisms, dominant land use (undisturbed forest in a state park, and potential controlling factors. An improved under- suburban, agricultural, urban residential, and heavily standing of the interactions between salinization urban commercial) (Table 1). All eight of these sites pulses and nutrient pollution is necessary to better are in close proximity and share the same hydrologic, manage streams and rivers and to identify unantici- geologic (piedmont), and biome (temperate decidu- pated geochemical relationships that may impact ous/humid). Seven of these eight sites are collocated stream restoration strategies and water quality goals. with US Geological Survey gaging stations (Table 1). Two of these eight sites (POBR, BARN) are nested in the relatively undisturbed 5500 ha Beaver Dam Methods Watershed, which drains into the Gunpowder River watershed (then Chesapeake Bay) in Northern Balti- We conducted laboratory experiments from sediments more County. The other 6 sites are nested in the collected at 12 sites across two metropolitan regions heavily modified suburban and urban Gwynns Falls and analyzed high-frequency sensor data from three Watershed, which drains 17,500 ha (and empties into sites to explore the potential relationships between the Patapsco River and then Chesapeake Bay) in episodic salinization and solute concentrations during Southern Baltimore County and Baltimore City snowstorms in streams. Laboratory experiments were (Fig. 1). used to characterize the potential release of base The remaining 4 of the 12 incubation stream sites cations, carbon, and nutrients during episodic are from the heavily urbanized Anacostia watershed 123 466 Biogeochemistry (2018) 141:463–486

Fig. 1 Maps of the study area, with different spatial scales for maps for Gwynns Falls and Baisman Run (Baltimore) water- each map. The high-frequency nitrate sensors are denoted as sheds and Anacostia (Washington DC) watershed are based on black squares in the Chesapeake Bay map (Rock Creek sensor is the National Land Cover Database 2011 (Homer et al. 2015) in Washington DC; Difficult Run sensor is in VA). Land use

(46,000 ha) in Southern Maryland, near Washington (27–41%) and offer a slightly different hydrology, DC. The Anacostia is a major tributary of the tidal geology (coastal plain), land use (Fig. 1), and drainage Potomac River and the Chesapeake Bay as described infrastructure (less leaky, less old, and more green/ in Smith and Kaushal (2015), and Devereux et al. Best Management Practices stormwater features) than (2010). These four sites vary in both drainage area the Baltimore LTER sites. Three of these four (200–18,800 ha) and impervious surface cover

123 Biogeochemistry (2018) 141:463–486 467

Table 1 Characteristics of the all sites for the sediment incubation experiments and the high-frequency sensor data analysis Site Land use Drainage area Impervious surface USGS Gaging Metropolitan Purpose context (ha) (%) Station area

POBR Forest 38 0 01583570 Baltimore, MD Incubation BARN Forest/suburban 382 \ 1 01583580 MCDN Forest/ 8 0 01589238 agriculture GFGL Suburban 81 19 01589180 GFGB Suburban 1065 15 01589197 GFVN Suburban 8349 17 01589300 DRKR Urban 1414 31 01589330 GRGF Urban 557 61 NA CC Urban 178 27 NA Washington, PB Urban 7925 32 01649190 DC NERP Urban 18777 29 01649500 Incubation/ Sensor SLIG Urban 1676 41 01650800 Incubation Difficult Suburban 15000 13 01646000 Sensor Run Rock Creek Urban 16600 32 01648010 Sensor

Anacostia watershed sites are collocated with US suburban and urban land cover, they also have a wide Geological Survey gaging stations (Table 1). and undeveloped riparian zone consisting of hiking Data from three sensors located in the Washington trails, recreational areas, and bands of intact forest DC Area were used in this study; Difficult Run, Rock with temperate deciduous trees (as described in Miller Creek, and Northeast Branch Anacostia. All three et al. 2013). The Difficult Run sensor (Gaging Station sensors are in close proximity to the Anacostia 01646000) is located near Great Falls, Virginia, Watershed incubation sites (Fig. 1), and are main- approximately 24 km northwest of Washington DC, tained and operated by the US Geological Survey. The and the Rock Creek sensor (gaging station 01648010) Northeast Branch Anacostia site was selected because is located in the northern corner of Washington DC the sensor is co-located with one of the experimental (Fig. 1). Difficult Run has a watershed size of incubation sites (NERP), and it has a 14 year contin- 15,000 ha, and Rock creek has a watershed size of uous record of high-frequency specific conductance 16,600 ha. Difficult Run has 5 years of continuous (15 min intervals) and discrete (grab-sample) chloride high-frequency 15 min interval measurements of data. The NERP sensor (Gaging Station 01649500) is specific conductance and nitrate, and Rock Creek a heavily urban site (Fig. 1) with short reaches of has a 2 year length of record, also in 15 min intervals. boulder embankment (previously channelized), and Both Difficult Run and Rock Creek are tributaries of has a watershed size of 18,800 ha. The NERP sensor the Potomac River and the Chesapeake Bay. Previous does not have high-frequency nitrate measuring studies have characterized the sediment and biogeo- capabilities. chemical dynamics in Difficult Run (e.g., Hupp et al. The Difficult Run and Rock Creek sites were 2013; Batson et al. 2015), and other work has selected due to the availability of high-frequency characterized nutrient loading and emerging contam- nitrate measurements, and because of similarities in inants in the Rock Creek watershed (e.g., Miller et al. their discharge and watershed size to sites where 2013; Battaglin et al. 2009). experiment incubation were performed (described above). Although both sites have a predominately

123 468 Biogeochemistry (2018) 141:463–486

Sediment incubation experiments across varying a shaking table (slow mode) in the dark for 24 h at degrees of salinization room temperature (20 °C). After the incubation, the water was immediately and carefully removed from Roughly 1 kg of sediment was collected from the the flask using a pipette as to avoid any disturbance to streambed per site using a clean shovel and a new the sediment, and then filtered through a pre-com- Ziploc bag during fall 2014. In order to achieve a busted Whatman 0.7 micron glass fiber filter. The representative sediment sample for each site, small filtered post-incubation water was stored in a fridge at amounts of sediment were gathered from three places 4 °C for water chemistry analysis (described below). (left bank, center, right bank) of two separate An aliquot of the post-incubation filtered water was transects, roughly 20 meters apart. Two liters of immediately acidified in a small acid-washed HDPE streamwater were also collected (via acid-washed Nalgene bottle to contain 0.5% high-purity nitric acid HPDE Nalgene bottles; no headspace). The sediment for base cation analysis and was stored at room and streamwater were transported in a chilled cooler to temperature for up to 12 months. Flask weights and a laboratory, and were kept cool and moist during the initial sediment weights were recorded prior to the experiment set-up. In order to homogenize the sample incubation. After the incubations, sediments were for particle size, the sediment was sieved in the lab dried in their flasks in a drying oven at 95 °C for 12 h, with a 2 mm sieve and the fine fraction (\ 2 mm) was then combusted in a furnace at 550 °C for 12 h. used for the incubation. Sixty grams of homogenized Sediment weights were recorded at every step and sediment were added to each acid-washed glass were used to calculate ash free dry mass to approx- Erlenmeyer flask along with 100 mL of unfiltered imate organic matter content. streamwater to simulate a vertical water column with a sediment–water interface. Sodium chloride was added Streamwater chemistry analyses to increase the salinity of the simulated stream columns at various treatments; 0, 0.5, 1, 2.5, 5, and Dissolved inorganic carbon (DIC), dissolved organic 10 g/L. This is a plausible range of salinity (0 to 6 g/L carbon (DOC, measured as non-purgeable organic chloride, 0 to 4 g/L of sodium), as long-term studies carbon), and total dissolved nitrogen (TDN) concen- have reported elevated measurements of both chloride trations in water were measured within 24 h after the (e.g. 8 g/L) and sodium (e.g. 3 g/L) during winter incubation using a combustion-catalytic-oxidation- months at the Baltimore sites (Kaushal et al. 2005; NDIR method on a Shimadzu Total Organic Carbon Kaushal et al. 2017); regression models have sug- Analyzer (TOC-V CPH/CPN; Shimadzu, Columbia, gested even higher concentrations of salinity (e.g. Maryland, USA). Soluble reactive phosphorus con- 14,000 lS/cm) following road salt applications at the centrations (SRP) in the water were measured within Anacostia sites (Miller et al. 2013). 10 days after the incubation using an automated In order to represent salt inputs to rivers (snowmelt colorimetric-blue method on a Lachat QuickChem with road salt), sodium chloride was dissolved into 8500 Series 2 FIA System (Hach, Loveland, Colorado, 100 mL unfiltered streamwater in a separate volumet- USA). Base cation (calcium, potassium, magnesium) ric flask before being pipetted onto sediment in the and Si concentrations in the acidified water samples Erlenmeyer flask. We acknowledge that salinization were measured within 12 months after the incubation can actually be a mixture of ions (sensu Kaushal et al. via inductively coupled plasma optical emission 2017; Kaushal et al. 2018; Kaushal et al. 2013), but spectrometry in an acidified analytical matrix on a used sodium chloride because it is commonly used as a Shimadzu Elemental Spectrometer (ICPE-9800; Shi- deicer. In order to isolate the sediment–water interac- madzu, Columbia, Maryland, USA). tion, a control flask of just unfiltered streamwater was also incubated along with the treatment flasks. All High-frequency sensors: specific conductance experiments for each site were incubated together in as a predictor for chloride and nitrate duplicates within 12 h of field collection. The flasks concentrations in streams were capped loosely with aluminum foil to limit evaporation but allow for air exchange to simulate High-frequency sensor data from US Geological open system conditions. The flasks were incubated on Survey stations 01649500 (NERP), 01646000 123 Biogeochemistry (2018) 141:463–486 469

(Difficult Run) and 01648010 (Rock Creek) (sites streamwater grab samples), the resulting dissolved described above) were analyzed to empirically char- concentrations were averaged, and the averaged acterize the in situ relationships between salinity values were used for all statistical analyses. In order (represented as specific conductance) and nitrate. We to isolate the effects of episodic salinization on the compared the potential relationships between salinity mobilization of base cations, carbon, and nutrients and nitrate from the incubation experiments with from the sediment to streamwater, the results from an sensor data from nearby sites. Specific conductance untreated control flask for each site were subtracted was measured using a submersible electrode sensor, from the results of each treatment (Duan and Kaushal calibrated to each site, and adjusted to represent the 2015). The resulting dissolved water chemistry were cross-sectional mean at the time of observation statistically evaluated using ordinary linear regres- (Radtke et al. 2005). At Difficult Run and Rock sions with sodium chloride treatment as the indepen- Creek, nitrate concentrations were measured using a dent variable and base cation, carbon, or nutrient Submersible Ultraviolet Nitrate Analyzer (SUNA, concentration as the dependent variable. Slopes with a Sea-Bird Scientific, Bellevue, Washington, USA) with p value \ 0.05 were assumed statistically significant, a 10 mm optical path length (Pellerin et al. 2013). The and this p-value criteria was used to ascertain whether nitrate optical sensors were lab calibrated to grab episodic salinization in laboratory experiments signif- samples from each site, and the optics were corrected icantly affected the dissolved concentrations of dif- for temperature and turbidity. Although the optical ferent solutes. The r2 coefficient of determination of sensor cannot distinguish between nitrate and nitrite, the linear regression model was calculated for each the measurement was assumed to be nitrate (as nitrite experiment, and was used to characterize the variabil- is negligible in these streams) (Pellerin et al. 2013). ity in dose responses (i.e., a high r2 value indicated The NERP sensor does not have nitrate measuring alignment of the solute response with the linear capabilities. Specific conductance (at all three sites) model). For the experimental incubations, all data and nitrate (at two sites) were measured by the sensors and statistical analyses was conducted using Microsoft in 15 min intervals, which were averaged into a single Excel (Microsoft Corporation, Redmond, Washing- daily value before analyses in this study. ton, USA). At all 3 sites, dissolved chloride was measured by Gaps in the high-frequency nitrate and specific ion chromatography in discrete grab samples collected conductance sensor measurements were not estimated approximately every 3 weeks by the US Geological by interpolation as they were normally distributed Survey. For the NERP site, 174 chloride measure- about their means, with the exception of some specific ments were available spanning roughly 13 years conductance outliers (e.g., the episodically high (2004–2017). For the Difficult Run site, 143 chloride specific conductance). At each site, a high-frequency measurements were available spanning roughly chloride record was estimated based on a linear 10 years (2007–2017). For the Rock Creek site, 259 relationship between the continuous high-frequency chloride measurements were available spanning specific conductance measurements and discrete chlo- roughly 5 years (2012–2017). At each site, a regres- ride grab sample measurements, as mentioned above. sion model was used to determine the relationship The magnitude of the relationship between estimated between precisely-concurrent measurements of dis- continuous chloride and sensor-measured continuous crete chloride concentrations (from grab samples) and nitrate was also assessed by ordinary linear regression the continuous high-frequency specific conductance with estimated chloride as the independent variable measurements (from sensors). From this relationship, and sensor-measured nitrate as the response variable. a high-frequency chloride (e.g., salinity) record was In order to assess the impact of extreme outliers, a least estimated based on the continuous high-frequency squares linear regression with iterative bi-square specific conductance measurements. weighting was also performed, and the resulting slope was compared to the ordinary linear regression slope Statistical analyses for agreement. A p-value \ 0.01 was used to demar- cate statistical significance of the slope, and whether For each site, the incubation experiments were con- an in situ empirical relationship exists between ducted in duplicates (using the same sediments and chloride and nitrate concentrations. The r2 coefficient 123 470 Biogeochemistry (2018) 141:463–486 of determination of the linear regression model was P at POBR to a maximum of 97.15 lg/L P at GFGL. used to assess the strength of the relationship. For the DIC varied from a minimum of 2.61 mg/L at POBR to high-frequency sensors component of this study, all a maximum of 38.70 mg/L at GFGL, while DOC data and statistical analyses were conducted using varied from a minimum of 1.45 mg/L at MCDN to a MATLAB (MathWorks Inc., Natick, Massachusetts, maximum of 4.52 mg/L at a small suburban creek USA). (CC, Anacostia). Silica concentrations varied from a minimum of 0.87 mg/L Si at a second order urban stream (NERP, Anacostia) to a maximum of 6.75 mg/ Results L Si at GFGL. Although elemental silicon was measured in this study, we are reporting the results Sediment incubation experiments across varying as silica as it is the dominant form of silicon in degrees of salinization streamwater (e.g., Treguer et al. 1995; Conley 2002). The sum of the base cations (calcium, magnesium, Base cations, nitrogen, silica, phosphorus and carbon potassium, and sodium) varied from a minimum of 9.19 mg/L at POBR to a maximum of 192.45 mg/L at As expected, the ambient streamwater chemistry a suburban mid-size stream with sanitary infrastruc- varied across sites (Table 2). Chloride varied from a ture leak issues (DRKR, Baltimore). The sediment minimum of 2.72 mg/L at a headwater forested site composition also varied between sites, with sediment (POBR, Baltimore) to a maximum of 135 mg/L at a organic matter ranging from a minimum of 0.17% at headwater suburban site (GFGL, Baltimore). TDN two mid-size streams (SLIG, Anacostia, suburban varied from a minimum of 0.33 mg/L N at a headwater with old growth trees) and (GRGF, Baltimore, urban forested site (POBR, Baltimore) to a maximum of and heavily polluted) to a maximum of 0.99% at 5.96 mg/L N at a small agricultural stream (MCDN, MCDN. Baltimore). SRP varied from a minimum of 2.05 lg/L

Table 2 Ambient site chemistry at the time of the sample collection (i.e., starting in-stream conditions for the sediment incubation experiments) Site Ambient stream conditions at time of sample grab TDN SRP DIC DOC Si Sediment Ca2? Mg2? K? Na? Cl- (mg/L) (lg/L) (mg/L) (mg/L) (mg/ OM (%) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) L)

POBR 0.33 2.05 2.61 2.23 1.49 0.46 0.44 1.80 1.90 5.05 2.72 BARN 1.49 3.50 4.37 1.71 3.68 0.24 8.25 4.40 2.40 15.45 40.6 MCDN 5.96 49.99 7.28 1.45 3.88 0.99 12.35 5.78 7.72 55.25 6.19 GFGL 1.09 97.15 38.70 2.89 6.75 0.70 75.20 39.50 5.15 23.90 135 GFGB 1.63 27.60 10.80 4.31 1.70 0.30 22.10 8.97 8.71 78.65 98.3 GFVN 1.69 31.35 18.21 2.53 2.00 0.35 38.85 20.65 10.02 59.25 51.6 DRKR 1.90 64.20 30.87 3.31 5.71 0.35 79.10 55.05 7.25 51.05 81.2 GRGF 2.18 12.20 29.27 4.49 4.48 0.17 65.60 21.05 7.95 54.90 89.7 CC 0.72 19.25 20.10 4.52 1.61 0.36 17.88 9.84 5.66 24.23 NA PB 1.50 24.30 7.32 1.77 1.72 0.22 19.35 8.57 8.16 53.35 63.2 NERP 0.86 12.90 11.96 4.47 0.87 0.44 26.25 11.85 12.05 68.10 31.0 SLIG 1.83 61.10 18.27 2.16 2.17 0.17 56.85 23.55 13.85 81.75 89.3 Broadly, the ambient chemistry was not a good predictor of the effects of episodic salinization. However, there were significant linear correlations between ambient SRP concentrations and the magnitude of SRP mobilization (i.e., incubation slopes), between ambient DIC concentrations and DIC response, and between ambient chloride concentrations and the magnitude of potassium mobilization

123 Biogeochemistry (2018) 141:463–486 471

There were consistent increases in base cation changes in cation concentrations. There were signif- concentrations with increasing salinity across sites icant linear increases in TDN concentrations with (Figs. 2, 3, 4). There were statistically significant increasing sodium chloride treatments at 9 of the 12 linear increases in calcium and potassium concentra- sites (Fig. 5), with rates of increase ranging from tions with increasing sodium chloride treatments at all 0.03 ± 0.001 to 0.13 ± 0.02 mg N per g NaCl. One 12 sites, with rates of increase ranging from of the sites had a statistically significant decrease in 0.20 ± 0.02 to 3.96 ± 0.37 mg Ca2? per g NaCl TDN concentrations with increasing salinity (BARN, and 0.38 ± 0.12 to 7.60 ± 0.53 mg K? per g NaCl, Baltimore, suburban), and salinity had no significant respectively. There were statistically significant linear effect on TDN at the remaining 2 sites (PB and SLIG, increases in magnesium concentrations at 11 of the 12 Anacostia, Urban). There was a significant decreasing sites with rates of increase ranging from 0.38 ± 0.12 linear response in silica concentrations to increasing to 2.35 ± 0.51 mg Mg2? per g NaCl. Salinization did sodium chloride treatments at only 1 of the 12 sites not have a significant effect on magnesium at the (POBR, Baltimore, forested). At 7 of the 12 sites, there remaining site (GFGB, Baltimore, suburban). Ambi- were statistically significant linear increases in SRP ent stream chemistry was not a good predictor of the concentrations with increasing sodium chloride treat- response of calcium or magnesium to salinization. ments (Fig. 6), with rates of increase ranging from However, there was a statistically significant linear 0.30 ± 0.08 to 5.63 ± 1.22 lg P per g NaCl correlation between ambient in-stream chloride con- (Table 4). Salinity had no statistically significant centrations during the sample grab, and the mobiliza- effect on SRP concentrations at the remaining five tion of potassium from the experiments (i.e., sites; however, all five of these sites responded with incubation rates of increase). increases in SRP concentrations in the treatment Increases in nutrient concentrations in response to experiments relative to the control experiment salinization were not as consistent across sites as (Fig. 6). Although the response of SRP concentrations

POBR (Forest) MCDN (Agriculture) BARN (Suburban) GFGB (Suburban)

r2=0.79 p<0.05 r2=0.95 p<0.05 r2=0.81 p<0.05 r2=0.86 p<0.05

GFVN (Suburban) GFGL (Suburban) CC (Urban) NERP (Urban)

r2=0.82 p<0.05 r2=0.84 p<0.05 r2=0.89 p<0.05 r2=0.97 p<0.05

DRKR (Urban) PB (Urban) SLIG (Urban) GRGF (Urban)

r2=0.93 p<0.05 r2=0.97 p<0.05 r2=0.96 p<0.05 r2=0.76 p<0.05

NaCl Addition (g/L)

Fig. 2 Responses of dissolved calcium to experimental salin- points, while the results from trial 2 are displayed as purple lines ization (relative to an untreated control incubation). There was a with triangular data points. The sites are ordered by watershed significant linear mobilization of calcium at all 12 sites. The impervious surface coverage from top left to bottom right, and results from trial 1 are displayed as gray lines with circular data the average of both trails was used to obtain the r2 and p-values

123 472 Biogeochemistry (2018) 141:463–486

POBR (Forest) MCDN (Agriculture) BARN (Suburban) GFGB (Suburban)

r2=0.97 p<0.05 r2=0.97 p<0.05 r2=0.90 p<0.05 r2=0.91 p<0.05

GFVN (Suburban) GFGL (Suburban) CC (Urban) NERP (Urban)

r2=0.97 p<0.05 r2=0.72 p<0.05 r2=0.94 p<0.05 r2=0.98 p<0.05

DRKR (Urban) PB (Urban) SLIG (Urban) GRGF (Urban)

r2=0.87 p<0.05 r2=0.92 p<0.05 r2=0.99 p<0.05 r2=0.96 p<0.05

NaCl Addition (g/L)

Fig. 3 Responses of dissolved potassium to experimental lines with triangular data points. The sites are ordered by salinization (relative to an untreated control incubation). There watershed impervious surface coverage from top left to bottom was a significant linear mobilization of potassium at all 12 sites. right, and the average of both trails was used to obtain the r2 and The results from trial 1 are displayed as gray lines with circular p-values data points, while the results from trial 2 are displayed as purple to episodic salinization was more variable than the sites (BARN and MCDN, Baltimore), there were response of TDN, the magnitude of the response was significant linear decreases in DIC concentrations with stronger (as indicated by the greater slopes in Table 4). increasing salinity with rates of change ranging from Broadly across all sites, the ambient stream chem- - 0.30 ± 0.07 to - 0.39 ± 0.08 mg C per g NaCl istry was not a good predictor of the response of (Table 4). The remaining 6 sites did not show a nitrogen or silica in the incubations (i.e., the presence significant increasing or decreasing trend for DIC; or strength of a salinization effect). However, there however, at most of these sites the DIC concentrations was a statistically significant weak linear correlation in the salinity treated experiments differed from their between ambient in-stream SRP concentrations during respective control experiments (Fig. 7). For the DOC sample grab and the strength of the response of SRP to response, there were statistically significant increasing experimental salinization (i.e., incubation rate of DOC concentrations with increasing sodium chloride change). The site with the lowest ambient SRP treatments at 3 of the 12 sites, with rates of increase concentration (POBR) had the weakest salinization ranging from 0.07 ± 0.02 to 0.19 ± 0.01 mg C per g effect (i.e., rate of SRP change), while the site with the NaCl. Although the remaining nine sites did not show highest ambient SRP concentration (GFGL) had the any significant trends for DOC, the concentrations strongest salinization effect. varied between treated experiments relative to the The response of carbon to episodic salinization control experiment, and between the individual salin- varied largely across sites and across salinity treatment ity treatment levels at each site (Fig. 7). For example, levels. There were statistically significant linear at a small agricultural stream (MCDN, Baltimore), increases in DIC concentrations with increasing there was an initial increase in DOC concentrations sodium chloride treatments at 4 of the 12 sites, with with the lower sodium chloride treatments, then a rates of change ranging from 0.06 ± 0.02 to decrease in DOC concentrations with the higher 0.32 ± 0.07 mg C per g NaCl (Table 4). At 2 of the salinization treatments (Fig. 7). Salinization did not 123 Biogeochemistry (2018) 141:463–486 473

POBR (Forest) MCDN (Agriculture) BARN (Suburban) GFGB (Suburban)

r2=0.80 p<0.05 r2=0.93 p<0.05 r2=0.77 p<0.05

GFVN (Suburban) GFGL (Suburban) CC (Urban) NERP (Urban)

r2=0.71 p<0.05 r2=0.93 p<0.05 r2=0.81 p<0.05 r2=0.92 p<0.05

DRKR (Urban) PB (Urban) SLIG (Urban) GRGF (Urban)

r2=0.72 p<0.05 r2=0.96 p<0.05 r2=0.94 p<0.05 r2=0.84 p<0.05

NaCl Addition (g/L)

Fig. 4 Responses of dissolved magnesium to experimental as purple lines with triangular data points. The sites are ordered salinization (relative to an untreated control incubation). There by watershed impervious surface coverage from top left to was a significant linear mobilization of magnesium at 11 of the bottom right, and the average of both trails was used to obtain 12 sites. The results from trial 1 are displayed as gray lines with the r2 and p-values circular data points, while the results from trial 2 are displayed have a detectable effect on sediment organic matter at directly ensuing a snow storm (e.g., road salt appli- any site (Supporting Fig. S4). cation). At each site, the episodically high specific There were no observable patterns between the conductance is at least an order of magnitude above ambient in-stream chemistry at a site at the time of both the baseflow measurements and the long-term sample grab and the response of DOC to salinization in average, and the magnitude and duration of the pulse this study. However, there was a statistically signif- varied between sites. There was a large variation in the icant positive linear correlation between the ambient chloride and nitrate data at both sensor sites. As in-stream DIC concentrations during sample grab, and expected, there was a strong positive correlation the response of DIC to salinization (i.e., incubation between specific conductance and chloride concentra- rate of change). Among the six incubation sites which tion at these sites (Fig. 8). There were weak but still showed a significant DIC response, the rate of change statistically significant increases in nitrate concentra- was negative (i.e., salinization removed DIC from tions with increasing chloride concentrations at both water column) when the starting ambient in-stream high-frequency sensor sites (Fig. 9). At the Difficult DIC concentration was low (\ 10 mg/L). At the sites Run sensor (near Great Falls, Virginia), the rate of with high ambient DIC concentrations during sample increase was 1.46 ± 0.12 mg N per g Cl (ANOVA, grab ([ 10 mg/L), experimental salinization had a r2 = 0.08, df = 1583, p \ 0.01), and at the Rock mobilization effect on DIC (Tables 2 and 4). Creek sensor (northern Washington DC), the rate of increase was 0.46 ± 0.06 mg N per g Cl (ANOVA, Relationships between high-frequency specific r2 = 0.10, df = 484, p \ 0.01). conductance and nitrate concentrations in streams

Episodic salinization is shown in Fig. 8 as a pulse of elevated specific conductance in the winter months

123 474 Biogeochemistry (2018) 141:463–486

POBR (Forest) MCDN (Agriculture) BARN (Suburban) GFGB (Suburban)

r2=0.77 p<0.05 r2=0.85 p<0.05 r2=0.99 p<0.05

GFVN (Suburban) GFGL (Suburban) CC (Urban) NERP (Urban)

r2=0.78 p<0.05 r2=0.96 p<0.05 r2=0.96 p<0.05 r2=0.93 p<0.05

DRKR (Urban) PB (Urban) SLIG (Urban) GRGF (Urban)

r2=0.89 p<0.05

r2=0.97 p<0.05

NaCl Addition (g/L)

Fig. 5 Responses of dissolved nitrogen to experimental salin- lines with triangular data points. The sites are ordered by ization for (relative to an untreated control incubation. There watershed impervious surface coverage from top left to bottom was a significant linear mobilization of TDN at 9 of the 12 sites. right, and the average of both trails was used to obtain the r2 and The results from trial 1 are displayed as green lines with circular p-values data points, while the results from trial 2 are displayed as red

Discussion scenario that is faced by streams and rivers experi- encing freshwater salinization syndrome (Fig. 8). Evidence for the water quality effects of episodic When it comes to the response of organisms to salinization and freshwater salinization syndrome salinization, recent research suggests that the extent of deviation from the norm (i.e., dosage and recovery Episodic salinization is a short-lived pulse in time) is more important that the duration of exposure, streamwater salinity and occurs during finite periods and may largely determine the adversity of its impact of time (i.e., several hours to several days) directly (Woo and Salice 2017). The same principal may be following a winter road salting event (i.e., winter extended to the geochemical and biogeochemical snowstorm). As with other chemical pulses, the stream behavior of a stream ecosystem upon salinization. salinity can increase by several orders of magnitude Episodic salinization can rapidly alter the pH and ionic during episodic salinization, which is an immense strength of streamwater, which can change the sedi- deviation from both the long-term norm and represents ment dispersal and coagulation dynamics, chemical a disturbance to ecosystem processes (e.g., Kaushal complexation reactions, and the microbial processing et al. 2014). In addition, salt concentrations during of nutrients (Kaushal et al. 2017; Kaushal et al. 2018). episodic salinization can exceed thresholds for sensi- Unlike at lower salinity levels, during episodically tive organisms (Kaushal et al. 2005). As such, episodic high salinity, cation exchange sites may become salinization can be a hot moment in biogeochemical saturated, biotic mechanisms might be inhibited, and cycling, and have the potential to permanently alter the there may be corresponding changes to nutrient and aquatic community and associated chemical process- carbon cycles, which can impact ecosystems (Duan ing (e.g., McClain et al. 2003). and Kaushal 2015). In our experiments, we attempted to simulate episodic salinization as it is a realistic exposure

123 Biogeochemistry (2018) 141:463–486 475

POBR (Forest) MCDN (Agriculture) BARN (Suburban) GFGB (Suburban)

r2=0.80 p<0.05 r2=0.76 p<0.05 r2=0.97 p<0.05

GFVN (Suburban) GFGL (Suburban) CC (Urban) NERP (Urban)

r2=0.95 p<0.05 r2=0.84 p<0.05 r2=0.90 p<0.05

DRKR (Urban) PB (Urban) SLIG (Urban) GRGF (Urban)

r2=0.85 p<0.05

NaCl Addition (g/L)

Fig. 6 Responses of dissolved phosphorus to experimental as red lines with triangular data points. The sites are ordered by salinization for (relative to an untreated control incubation). watershed impervious surface coverage from top left to bottom There was a significant linear mobilization of SRP at 7 of the 12 right, and the average of both trails was used to obtain the r2 and sites. The results from trial 1 are displayed as green lines with p-values circular data points, while the results from trial 2 are displayed

BARN (Suburban) MCDN (Agriculture) NERP (Urban)

r2=0.81 p<0.05 r2=0.85 p<0.05

r2=0.84 p<0.05

NaCl Addition (g/L)

Fig. 7 Responses of DIC and DOC to experimental salinization value indicates carbon was removed from the water column. The for 3 of the 12 sites. The effects of salinization on carbon was results from trial 1 are displayed as blue lines with circular data variable, and this subset of sites was selected to demonstrate the points, while the results from trial 2 are displayed as orange lines variability. For each treatment level, a positive value indicates with triangular data points, and the average of both trails was carbon was released from the sediment into the water column used to obtain the r2 and p-values relative to an untreated control incubation, while a negative

123 476 Biogeochemistry (2018) 141:463–486

7000 700 Northeast Branch Anacostia, MD (NERP) R2 = 0.975, n=174 6000 DA = 18800 ha 600 y=0.31x-32.11 5000 500

4000 400

3000 300

2000 Chloride (mg/L) 200

1000

Specific Conductance (µS/cm) 100

0 0 0 500 1000 1500 2000 2500 2006-01 2008-01 2010-01 2012-01 2016-01 2004-01 2014-01 2018-01 Specific Conductance (µS/cm)

8000 1600 Rock Creek, Washington DC 7000 1400 R2 = 0.978, n=259 DA = 16600 ha 6000 1200 y=0.29x-30.80

5000 1000

4000 800 3000 600

2000 Chloride (mg/L) 400

Specific Conductance (µS/cm) 1000 200 0 0 0 1000 2000 3000 4000 5000 6000 2013-01 2016-01 2017-01 2014-01 2015-01 Specific Conductance (µS/cm) 3500 1000 Difficult Run, VA 3000 R2 = 0.992, n=143 DA = 15000 ha 800 y=0.32x-29.08 2500

2000 600

1500 400 1000 Chloride (mg/L)

500 200 Specific Conductance (µS/cm)

0 0 0 500 1000 1500 2000 2500 3000 2013-01 2017-01 2007-01 2009-01 2011-01 2015-01 Specific Conductance (µS/cm)

Fig. 8 (Left panel) Timeseries of daily average values of strong, positive, linear relationship between specific conduc- specific conductance. Episodic salinization manifests in the tance and chloride concentrations. Dotted red line is the 95% high-frequency sensor record as a pulse in specific conductance confidence bounds for the regression. NERP is also an in the hours to days following a snowstorm. Experiments were incubation experiment site. Difficult Run and Rock Creek are conducted to determine potential effects of the salinization pulse nearby to the incubation experiment sites, and the in situ on sediment biogeochemistry. (Right panel) Relationship relationships between salinity and nitrate were assessed at these between dissolved chloride from grab samples and the nearest sites instantaneous specific conductance measurement. There was a

123 Biogeochemistry (2018) 141:463–486 477

1.6 Our results are consistent with a previous study

1.4 evaluating the effect of experimental salinity on wetland sediment biogeochemistry, which reported 1.2 increased levels of magnesium, potassium, sodium,

1 calcium, and decreased pH with long-term salinization (Kim and Koretsky 2013). The authors concluded that 0.8 ion exchange would result in a greater release of Nitrate (mg/L as N) Rock Creek, Washington DC 0.6 cations with a greater salinity treatment, and that pH y=0.46x + 0.83, r2=0.10 n=486 was suppressed due to the interactive effects of 0.4 carbonate precipitation, oxidation, and ion exchange 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 reactions (which would affect iron and manganese Chloride (g/L) cycling). We suggest similar mechanisms in our 3 results, although we evaluated the potential effects

2.5 of episodic salinization on stream water quality. Positive relationships between salinization and 2 base cations have also been demonstrated in the soil literature. Column leaching experiments of salinizing 1.5 urban and forested soils in Sweden and central USA 1 have indicated that sodium preferentially displaces Difficult Run, VA Nitrate (mg/L as N) 2 calcium, potassium, and magnesium during ion 0.5 y=1.46x + 1.43, r =0.09 n=1585 exchange reactions, leading to the mobilization of 0 these base cations (Norrstrom and Bergstedt 2001; 0 0.2 0.4 0.6 0.8 1 Lofgren 2001; Robinson et al. 2017). Polyvalent Chloride (g/L) elements, such as calcium and magnesium were 18–51 Fig. 9 Linear relationships between chloride (as estimated times higher in the salt solution leachate, while the from daily average values of specific conductance) and nitrate in monovalent cations, such as potassium and sodium, the high-frequency sensor record. The slope from the Rock were 2–6 times higher in the salt solution leachate Creek sensor (top panel) is within the same order of magnitude (Lofgren 2001). Furthermore, the increased ionic as the mean slope from the experiments, while the slope from the Difficult Run sensor (bottom panel) is an order of magnitude strength caused by chloride was found to easily larger (contrasted in units of mg N per g Cl). The NERP sensor release hydrogen ions leading to an initial decrease lacks nitrate measuring capabilities in the pH of the soil. Decreases in pH can affect the edge charge on clay minerals, and at high salinity Potential effects of episodic salinization levels, can cause organic matter-cation colloid disper- and freshwater salinization syndrome on base cations sal (Norrstrom and Bergstedt 2001). At lower salt concentrations, a suppressed pH can stabilize organic In the incubation experiments, there were consistent matter-cation colloid dispersal, however, limiting the mobilizations of base cations from the sediment into mobilization of base cations (Norrstrom and Bergstedt the water column in response to episodic salinization 2001). Surface soils receiving salt applications have (Figs. 2, 3, 4). Rates of increase for calcium and been shown to be enhanced in sodium and depleted in potassium were greater in the Anacostia watershed calcium relative to their surroundings when subjected than in the Baltimore watersheds (Table 3), which to sodium chloride road salting, or enhanced in could be attributed to differences in the underlying magnesium and depleted in sodium when subjected lithology and sediment as most of the Anacostia to magnesium chloride road salting – both of which are watershed is within the Atlantic Coastal Plain phys- further evidence for ion exchange, complexation or iographic region, while the Baltimore watersheds are sorption of organic matter, or dispersal of colloids within the Piedmont physiographic region. However, (Backstrom et al. 2004; Cunningham et al. 2008). ambient stream chemistry was not a good predictor of The mobilization of base cations from soils and the response of base cations to salinization. sediments to streamwater due to episodic salinization and freshwater salinization syndrome could further 123 478 Biogeochemistry (2018) 141:463–486

Table 3 Results of a linear regression analysis of the base cation response at each incubation site to sodium chloride additions Site Ca2? K? Mg2? Slope p-value r2 Slope p-value r2 Slope p-value r2

POBR 1.377 – 0.356 0.018 0.79 1.054 – 0.093 3.E-04 0.97 1.675 – 0.424 0.017 0.80 BARN 2.840 – 0.697 0.015 0.81 1.726 – 0.291 0.004 0.90 2.228 – 0.615 0.022 0.77 MCDN 0.199 – 0.023 0.001 0.95 0.408 – 0.037 4.E-04 0.97 0.589 – 0.081 0.002 0.93 GFGL 2.446 – 0.542 0.011 0.84 0.376 – 0.118 0.033 0.72 1.311 – 0.179 0.002 0.93 GFGB 1.494 – 0.296 0.007 0.86 1.517 – 0.239 0.003 0.91 0.060 ± 0.131 0.673 0.05 GFVN 0.338 – 0.080 0.014 0.82 0.469 – 0.038 2.E-04 0.97 0.382 – 0.122 0.035 0.71 DRKR 1.162 – 0.154 0.002 0.93 2.149 – 0.415 0.007 0.87 0.564 – 0.176 0.033 0.72 GRGF 3.117 – 0.873 0.023 0.76 5.665 – 0.583 0.001 0.96 2.352 – 0.508 0.010 0.84 CC 1.575 – 0.280 0.005 0.89 2.483 – 0.309 0.001 0.94 0.529 – 0.129 0.015 0.81 PB 2.038 – 0.170 3.E-04 0.97 4.748 – 0.721 0.003 0.92 0.541 – 0.057 0.001 0.96 NERP 3.964 – 0.366 4.E-04 0.97 7.597 – 0.535 1.E-04 0.98 1.169 – 0.176 0.003 0.92 SLIG 2.498 – 0.253 0.001 0.96 5.367 – 0.205 1.E-05 0.99 0.848 – 0.111 0.002 0.94 Units for calcium slopes are mg-Ca per g-NaCl, potassium slopes are mg–K per g-NaCl, and magnesium slopes are mg-Mg per g-NaCl. Bold slopes are significant at the p \ 0.05 level enhance the base cation loads of streams and rivers Potential effects of episodic salinization (Norrstrom and Bergstedt 2001; Robinson et al. 2017). and freshwater salinization syndrome on dissolved In our study, the increases in base cations could be due carbon to the mobile anion effect as the presence of anions (i.e., nitrate, chloride) enable the leaching of base The response of carbon to episodic salinization varied cations from exchange sites on the sediment into the largely across sites and across salinity treatment levels water column, and we suggest this could increase the (Fig. 7). Our results agree with a recent sediment base cation loads of streams and rivers. This is salinization study, which indicated that salinization consistent with stream monitoring work in small can increase sediment releases of DOC at some sites forested watersheds showing 14% of the magnesium but not others potentially based on organic matter and 19% of the calcium flux (i.e., roughly a 20% quality (Duan and Kaushal 2015). Although our study increase) was attributed to mobilizations induced by did not characterize the quality of DOC, Duan and salt inputs (Price and Szymanski 2014). In stream Kaushal (2015) found that salinization consistently monitoring studies of medium sized watersheds, road and considerably increased net releases of the protein- salt inputs increased the net calcium and magnesium like fluorophore, but the effects of salinization on flux each by 44% above respective background, humic-like fluorophore releases were not consistent. increased the total solute flux in streamflow by This difference was attributed to variations in the 120%, and caused a net flushing (e.g., loss) of sodium effect of salinization on humic versus protein fractions from the watershed (Shanley 1994). Finally in long- of organic carbon. Increasing ionic strength caused by term monitoring studies, Kaushal et al. (2017) mea- salinization can enhance the solubility of the proteina- sured long-term increases in base cation concentra- ceous materials via sodium dispersion processes tions at some of these same sites in our study (which (Green et al. 2008, 2009) or through nonspecific were on average up to 60 times greater than a nearby electrostatic interactions at low salinities (e.g., a forested reference), which suggested mobilization of ‘‘salting-in’’ effect) (Tanford 1961; Chen et al. 1989). base cations by salinization. Humic fractions of organic carbon, which are larger hydrophobic molecules that occur in the colloidal size range (e.g., Aiken et al. 1985), can experience an array of effects when subjected to increasing ionic strength

123 Biogeochemistry (2018) 141:463–486 479 or salinization. For example, humic molecules can sediment (Table 4). These sites were all suburban flocculate (e.g., Sholkovitz 1976) or undergo sorption and urban sites. Here, we can still invoke the same to mineral surfaces (Fox 1991; Hedges and Keil 1999), conceptual model and biogeochemical mechanisms depending on the magnitude of the pH suppression proposed by Duan and Kaushal (2015) to interpret the caused by salinity (Kipton et al. 1992; Li et al. positive salinization effect on DOC release. That is, 2007, 2013). Because stream sediments are generally organic matter at these suburban and urban sites was enriched in labile, proteinaceous materials that are probably more labile (containing more protein-like derived from biofilms (algae and microbes), or components) and salinization consistently and con- wastewater organics in urban watersheds (e.g., Daniel siderably increased net releases of the protein-like et al. 2002), it is reasonable that salinization has fluorophore (Duan and Kaushal 2015). At other sites, potential to mobilize a significant amount of DOC the effect of salinization on DOC varied within these from some stream sediments. salinity ranges, probably reflecting combined effects Amrhein et al. (1992) also argued that decreases in that we have discussed above on humic substances or pH and increases in ionic strength of the streamwater, total DOC. For example, at a few sites (DRKR, GRGF, induced by salinization, could pull DOC-metal col- BARN), we found that salinization increased DOC loid complexes from the benthic surface into the water concentrations at the lower salinity treatments, but column where they could remain suspended due to then decreased concentrations at the higher salinity their interaction with the other charged particles or due treatments (Fig. 7 and Supporting Fig. S2) Sediments the hydrology. Steele and Aitkenhead-Peterson (2012) from all 12 sites had an estimated organic matter suggest that the solubility of DOC and other nutrients content of less than 1% at the time of sample collection substantially increases with increased salinity (sodium (Table 2), however, the extent of prior salt exposure of absorption ratio) in urban soils throughout the state of the sediment at each site could also impact the Texas, USA. However, Compton and Church (2011) response of DOC (Green et al. 2009). conducted short and long-term soil incubations with For DIC, the response to episodic salinization also low salinity treatment levels and concluded that varied largely across sites. The effects of episodic increasing salinity removes DOC from solution via salinization on DIC releases from sediments to flocculation and sorption to mineral surfaces, and thus streamwater may involve changes in the dissolution prevents its leaching to streams. As a result of of carbonate minerals, or changes in the organic variations in organic matter content and sorption carbon mineralization process and subsequent carbon capacity across sites, the effect of salinization on DOC dioxide efflux, which may be coupled with DOC releases varies but tends to be driven by a combination biogeochemistry. On one hand, the solubility of of the array of mechanisms mentioned above (Am- carbonate minerals increases with salinization (Akin rhein et al. 1992; Evans et al. 1998; Green et al. and Lagerwerff 1965). On the other hand, the solubil- 2008, 2009; Compton and Church 2011; Ondrasek ity coefficient for carbon dioxide decreases with et al. 2012). salinity (Weiss 1974; Duan and Sun 2003). Duan Relative to prior sediment salinization experiments, and Kaushal (2015) observed that salinization our study examined the effects of salinization across increased DIC releases from sediment at lower salinity wider ranges; 0, 0.5, 1.0, 2.5, 5.0, and 10.0 g/L sodium levels due to the dominance of the former effect, but chloride (i.e., 0 to 6 g/L chloride). Some studies decreased DIC releases at higher salinity levels due to suggest that there may be a specific threshold salinity the dominance of the latter effect. Duan and Kaushal level between which salts stabilize colloids (low (2015) reported sites showing a consistent positive salinity) or mobilize colloids (high salinity) (Norr- response of DIC to salinization (increasing DIC strom and Bergstedt 2001; Green et al. 2008; Kim and concentration) or a consistent negative response Koretsky 2013). We found that the effects of salin- (decreasing DIC concentration). Although our study ization on DOC release from sediment are more had a wider range of salinity treatments, our results are complicated across larger salinization ranges. For consistent with Duan and Kaushal (2015) and their example, at 5 of our 12 sites (suburban and urban: conceptual model as DIC concentrations increased in GFGL, GFGB, GFVN, NERP, CC), salinization 4 of the 12 incubation sites and decreased in two sites consistently increased net releases of DOC from (Table 4). Moreover, elevated salinity for long periods 123 480 123

Table 4 Results of a linear regression analysis of the nutrient response at each incubation site to sodium chloride additions Site TDN SRP DIC DOC Si Slope p- r2 Slope p- r2 Slope p- r2 Slope p- r2 Slope p- r2 value value value value value

POBR 0.049 – 0.014 0.023 0.77 0.087 ± 0.036 0.074 0.59 0.048 ± 0.032 0.200 0.37 0.013 ± 0.030 0.698 0.04 2 0.012 – 0.004 0.034 0.71 BARN - 0.074 ± 0.021 0.026 0.75 0.300 – 0.084 0.023 0.76 2 0.301 – 0.072 0.014 0.81 - 0.029 ± 0.025 0.317 0.25 0.043 ± 0.040 0.344 0.22 MCDN 0.065 – 0.014 0.009 0.85 1.346 – 0.333 0.016 0.80 2 0.395 – 0.083 0.009 0.85 - 0.007 ± 0.037 0.864 0.01 - 0.078 ± 0.068 0.318 0.25 GFGL 0.051 – 0.005 0.001 0.96 5.634 – 1.221 0.010 0.84 0.323 – 0.073 0.011 0.83 0.069 – 0.024 0.042 0.69 0.031 ± 0.041 0.499 0.12 GFGB 0.026 – 0.001 3.E-05 0.99 2.1082 – 0.195 4.E-04 0.97 0.158 – 0.042 0.019 0.78 0.190 – 0.013 0.000 0.98 0.008 ± 0.014 0.588 0.08 GFVN 0.043 – 0.011 0.019 0.78 1.752 – 0.201 0.001 0.95 0.097 ± 0.073 0.256 0.31 0.055 ± 0.023 0.077 0.58 - 0.039 ± 0.018 0.095 0.54 DRKR 0.127 – 0.012 4.E-04 0.97 1.074 – 0.227 0.009 0.85 0.282 – 0.097 0.044 0.68 - 0.006 ± 0.017 0.720 0.04 - 0.003 ± 0.026 0.903 0.01 GRGF 0.070 – 0.012 0.004 0.89 1.491 ± 0.712 0.104 0.52 - 0.012 ± 0.026 0.681 0.05 - 0.024 ± 0.010 0.067 0.61 0.015 ± 0.008 0.130 0.48 CC 0.029 – 0.003 0.001 0.96 4.180 – 0.712 0.004 0.90 - 0.005 ± 0.037 0.890 0.01 0.075 ± 0.030 0.069 0.60 0.013 ± 0.005 0.068 0.61 PB 0.021 ± 0.031 0.537 0.10 0.329 ± 0.405 0.462 0.14 0.032 ± 0.173 0.862 0.01 0.041 ± 0.059 0.527 0.11 0.012 ± 0.008 0.183 0.39 NERP 0.131 – 0.018 0.002 0.93 0.106 ± 0.147 0.509 0.12 0.113 ± 0.059 0.127 0.48 0.071 – 0.016 0.010 0.84 0.012 ± 0.005 0.075 0.59 SLIG 0.005 ± 0.006 0.404 0.18 2.424 ± 2.256 0.343 0.22 0.065 – 0.019 0.029 0.74 0.049 ± 0.028 0.159 0.43 0.017 ± 0.007 0.071 0.60

Units for TDN slopes are mg-N per g-NaCl, SRP slopes are lg-P per g-NaCl, DIC and DOC slopes are mg-C per g-NaCl, and silica slopes are mg-Si per g-NaCl. Bold slopes are 141:463–486 (2018) Biogeochemistry significant at the p \ 0.05 level Biogeochemistry (2018) 141:463–486 481 may suppress microbial transformations of carbon in salinization has been shown to increase net mineral- the sediment and pore-water, which could alter in- ization in both roadside soils (Compton and Church stream DIC concentrations (Oren 2001). Similar to 2011; Green et al. 2008) and forested debris dams in a DOC, our study showed that the effect of salinization lab environment (Hale and Groffman 2006), which on DIC dynamics is more complicated, if we extend could further contribute to elevated ammonium con- salinity to a wider and more realistic range of centrations. Finally, several previous studies have also salinization for urban sites. shown appreciable leaching of dissolved organic nitrogen (DON; another important TDN species) from Potential effects of episodic salinization plant litter or soils along with DOC upon increased and freshwater salinization syndrome on dissolved salinization (Steele and Aitkenhead-Peterson 2013; nitrogen, phosphorus, and silica Green et al. 2009). Unlike ammonium or DON, the effects of saliniza- The variability in the response of TDN concentrations tion on nitrate fluxes varied considerably in previous to salinization between the Baltimore sites (where studies, potentially due to the competing processes of 88% of sites exhibited a positive response) and nitrification (due to the enhanced ammonium avail- Anacostia watershed sites (where 50% of sites exhib- ability) and nitrate removal via denitrification. Duck- ited a positive response) could be due to the sediment worth and Cresser (1991) concluded that salinization conditions and stream concentrations during the has no effect on nitrate release in forested soils, sample grab. The two sites that did not show a however, these soils were highly organic. Although significant effect of salinity in the Anacostia water- Compton and Church (2011) reported increases in shed had nearly twice the ambient TDN concentra- nitrogen mineralization in response to salinity, they tions than the other two Anacostia sites. This discuss how the extent of nitrification (or immobiliza- potentially indicated that the sediment at these two tion) of the released ammonium depends heavily on sites is already low in nitrogen content, perhaps due to the response of DOC to salinity, and thus the prior leaching from prior exposure to chloride (as implications for nitrate concentrations are unclear. discussed in Hale and Groffman 2006). Broadly across Duan and Kaushal (2015) reported net nitrate retention all sites, the ambient stream chemistry was not a good with increasing salinization in sediments and net predictor of the response of nitrogen in the incubations nitrate release with salinization in soils, after con- (i.e., the presence or strength of a salinization effect on ducting experiments from a subset of the exact sites nitrogen). used in this study. However, they also identified Our results of episodic salinization affecting nitro- ammonium and DON to be the dominant sources of gen dynamics are consistent with previous salinization nitrogen that were leached out of sediments or soil studies of stream sediments (Duan and Kaushal 2015), during the experiments, which could account for our wetland sediments (Kim and Koretsky 2013), and soils observed result of increases in TDN concentrations (Green et al. 2008; Compton and Church 2011; with salinization. Interestingly, at some of the same Duckworth and Cresser 1991). Similarly, our results sites as this study there were consistent positive could be interpreted within the context of these relationships between in-stream nitrate and chloride studies. First, ammonium, an important TDN species, concentrations in the long-term stream chemistry data can be mobilized by salinization due to sodium at the Baltimore LTER site (Kaushal et al. 2017). dispersion and ion exchange. (Duckworth and Cresser The results of our incubation experiments can be 1991; Green et al. 2008; Compton and Church 2011; used to interpret the significant linear relationships Kim and Koretsky 2011). As a positively charged ion, between in situ nitrate and chloride at two high- ammonium can easily be adsorbed on negatively frequency sensors sites. The mean rate of increase for charged particles of soils and sediments (Nieder et al. the nine experiments that showed a statistically 2011). Once retained on the sediment particle’s cation significant increase in TDN concentrations with exchange sites, ammonium can exchange with sodium increasing salinization was 0.11 mg N per g Cl (with ions, which could cause flushing of ammonium a standard deviation of 0.06). The mean slope for the induced by salinization (Duckworth and Cresser experiments is within the same order of magnitude as 1991; Kim and Koretsky 2011). Furthermore, the slope from the Rock Creek high-frequency sensor, 123 482 Biogeochemistry (2018) 141:463–486 and an order of magnitude smaller than the slope from (Bunn et al. 2002; Saiers and Lenhart 2003), both of the Difficult Run high-frequency sensor. which can be caused by salinization (Green et al. Despite the statistically significant positive rela- 2008). Iron complexes in colloidal organic matter can tionship between specific conductance (or chloride) bind phosphorus from the water column. The reduc- and nitrate concentrations, the relationship was weak, tion of SRP with salinization could also be due to the goodness of fit is low, and the model has no biotic perturbations, such as the temporary inhibition predictive power. There are many hydrological and of microbial activity at higher salinity or the rapid biogeochemical factors that mask the in situ response uptake of any released phosphate (Srividya et al. 2009; of nitrate concentration to increases in salinity, and it Oren 2001). is unlikely that the underlying factors in the experi- Our results of increasing SRP with salinization are mental incubations would be exactly the same as a more consistent with investigations of stream, wet- stream channel. For example, the type of salt additions land, and marine sediments. Although Duan and used in our experiment (lab grade sodium chloride) is Kaushal (2015) determined SRP release from stream not the same type of salt-sand mixtures used on sediments to decrease with salinization at five of the pavements (magnesium chloride, calcium chloride, same sites in this study, they also reported that releases potassium chloride, organic deicers). There are also of SRP from sediments increased at two urban sites. differences in the relative abundance of nitrogen They suggested that these SRP increases were coupled species in the incubation experiments and the stream with the release of large amounts of labile DOC, which (only TDN was measured in our experiments, only could result in redox conditions that are favorable for in situ nitrate was measured by the sensor). SRP release. However, in our study, there was no Salinization may cause increases in TDN (includ- observable relationship between the responses of SRP ing nitrate) in the stream, potentially due to the and DOC. Furthermore, increased salinity has been mobilization of nitrogen from sediments to stream associated with increases in streamwater phosphate water, coupled with enhancements in net nitrification, concentrations and net watershed phosphorus export or due to the leaching of nitrogen from soils through- in a recent study (Merrikhpour and Jalali 2012). out the watershed and the subsequent flushing to Experimental studies of wetland sediments indicated streams during storm flow. Although urban watersheds both releases of phosphate with salinization, attributed are hydrologically connected to impervious surfaces to enhanced anaerobic organic matter (and phospho- (Kaushal and Belt 2012), a major nitrogen source is rus) mineralization (Kim and Koretsky 2013), or the sewage leaks (Pennino et al. 2016; Newcomer-John- retention of phosphorus on sediments with saliniza- son et al. 2014; Kaushal et al. 2011), and so it is more tion, attributed to the formation and precipitation of likely that the nitrate is generated from flow-path insoluble iron-phosphate complexes (Baldwin et al. biogeochemical processes (e.g., nitrification) than 2006). In marine and estuarine sediments, phosphate from being washed-in into the channel during snow- was mobilized (i.e., decreased sorption) linearly with storms. Furthermore, because both nitrate and chloride increasing salinity, and the magnitude of release concentrations in these watersheds can be higher in depended heavily upon the sediment’s exchangeable groundwater than streamwater (Mayer et al. 2010; phosphorus content (Zhang and Huang 2011; Clavero Kaushal et al. 2005), it is also possible that the in situ et al. 1990; Spiteri et al. 2008). These releases of positive relationships are due to nitrogen and chloride phosphorus in coastal environments were attributed to being derived from the same groundwater hotspot in a combination of desorption from iron and aluminum the riparian zone near the sensors. oxides with increasing salinity, increased aqueous The mobilization of SRP from sediments in this complexation of phosphate with cations, and enhance- study seems to contradict previous studies in soils ments in anoxic microbial sulfate reduction (Spiteri showing that salinization reduces SRP concentrations et al. 2008). Changes in SRP dynamics between (Jun et al. 2013; Compton and Church 2011; Duan and sediments and streamwater with salinization are Kaushal 2015). The reduction of SRP upon saliniza- complex and warrant further evaluation, especially tion can be attributed to the reduced stability of within the context of redox changes. colloidal humic iron-aluminum phosphate complexes Relative to nitrogen or phosphorus, the effects of with increasing ionic strength and decreasing pH salinization on silica transformations are relatively 123 Biogeochemistry (2018) 141:463–486 483 less known and unstudied. There were no clear trends been updated since 1988 (Corsi et al. 2015). Dissolved in silica concentrations between salinity treatments or salts in freshwater are currently not regulated at any sites in our experimental incubations. Previous geo- broad regional scale (i.e., federal or state) for drinking chemical studies have reported that sodium chloride water in the U.S. (Kaushal 2016), possibly because causes a release of silicate from quartz particles (e.g., there has been little information on the long-term Dove and Elston 1992), potentially indicating that effects of salinization on water quality, environmental silica could be released with salinization in quartz-rich health, or human well-being (Canedo-Arguelles et al. soils and sediments. Some wetland sediments have 2016). Given growing evidence demonstrating mobi- also shown to release silicate upon experimental lization of nutrients and other ions, managing the salinity intrusion (e.g., Weston et al. 2006). However, effects of freshwater salinization syndrome and our results suggest that silica is much more stable un- episodic salinization represents a serious emerging der enhanced salinity in stream sediments. This issue for improving the efficacy of water quality stability is consistent with the relatively conservative management and stream and river restoration projects. behavior of silica across salinity gradients in many estuarine environments. The response of silica to Acknowledgements Steven Hohman, Jonathan Coplin, salinization may be of interest as silica can alleviate Benjamin Smith, and Rose Smith assisted with experimental set up and laboratory analyses. Nathan Bailey, Kelsey Wood, salinity stress and salt toxicity in a wide range of plants and Tom Doody provided logistical support. We also thank the (e.g., Romero-Aranda et al. 2006; Shi et al. 2013). USGS Richmond Field Office and USGS MD-DE-DC Water Thus, a release of silica on a watershed scale in Science Center for the maintenance of stream sensors and response to salinization could potentially limit the providing access to data. Funding was provided by NSF EAR 1521224, NSF SEES 1426844 and NSF DEB 1027188. ecosystem impacts of episodic salinization. Nonethe- less, the effect on silica warrants further study and investigation. References

Implications for managing episodic salinization Aiken GR, McKnight DM, Wershaw RL et al (1985) Human and freshwater salinization syndrome substances in soil, sediment, and water: geochemistry, isolation, and characterization. Wiley, New York Akin GW, Lagerwerff JV (1965) Calcium carbonate equilibria Increases in the frequency and magnitude of episodic in aqueous solutions open to the air. I. The solubility of salinization of rivers and streams is expected to occur calcite in relation to ionic strength. Geochim Cosmochim with future urbanization, given the strong relationship Acta 29(4):343–352 between impervious surface cover and salinization of Amrhein C, Strong JE, Mosher PA (1992) Effect of deicing salts on metal and organic matter mobilization in roadside soils. fresh water (Kaushal et al. 2005; Dugan et al. 2017; Environ Sci Technol 26(4):703–709 Gutchess et al. 2018). The mobilization of nutrients Backstrom M, Karlsson S, Backman L et al (2004) Mobilisation caused by episodic salinization, as evidenced in this of heavy metals by deicing salts in a roadside environment. study, could undermine costly efforts to curtail Water Res 38(3):720–732 Baldwin DS, Rees GN, Mitchell AM et al (2006) The short-term watershed nutrient pollution. Perturbations in the effects of salinization on anaerobic nutrient cycling and watershed carbon cycle, could cause long-term microbial community structure in sediment from a fresh- changes in organic matter lability, aquatic food webs, water wetland. Wetlands 26(2):455–464 and microbial and plant communities. Changes in Barnes RT, Raymond PA (2009) The contribution of agricul- tural and urban activities to inorganic carbon fluxes within ionic strength represented by increases in multiple temperate watersheds. Chem Geol 266(3–4):318–327 ions from sediments and soils as symptoms of Batson J, Noe GB, Hupp CR et al (2015) Soil greenhouse gas freshwater salinization syndrome could also impact emissions and carbon budgeting in a short-hydroperiod contaminant toxicity. Therefore, watershed manage- floodplain wetland. J Geophys Res Biogeosci 120(1):77–95 ment plans should now consider targeted strategies for Battaglin WA, Rice KC, Focazio MJ et al (2009) The occurrence reducing salt pollution and episodic salinization. The of glyphosate, atrazine, and other pesticides in vernal pools current US Environmental Protection Agency water and adjacent streams in Washington, DC, Maryland, Iowa, quality criteria for chloride, 4 day (chronic) average and Wyoming, 2005–2006. Environ Monit Assess 155(1–4):281–307 concentration exceeding 230 mg/L or 1 h (acute) average concentration exceeding 860 mg/L, has not 123 484 Biogeochemistry (2018) 141:463–486

Bernhardt ES, Palmer MA, Allan JD et al (2005) Synthesizing 273 to 533 K and from 0 to 2000 bar. Chem Geol U.S. River restoration efforts. Science 308(5722):636–637 193(3–4):257–271 Bunn RA, Magelky RD, Ryan JN et al (2002) Mobilization of Duan S, Kaushal SS, Groffman PM et al (2012) Phosphorus natural colloids from an iron oxide-coated sand aquifer: export across an urban to rural gradient in the Chesapeake effect of pH and ionic strength. Environ Sci Technol Bay watershed. J Geophys Res Biogeosci 36(3):314–322 117(G1):G01025. https://doi.org/10.1029/2011JG001782 Canedo-Arguelles M, Kefford BJ, Piscart C et al (2013) Salin- Duckworth CMS, Cresser MS (1991) Factors influencing isation of rivers: an urgent ecological issue. Environ Pollut nitrogen retention in forest soils. Environ Pollut 72(1):1–21 173:157–167 Dugan HA, Bartlett SL, Burke SM et al (2017) Salting our Canedo-Arguelles M, Hawkins CP, Kefford BJ et al (2016) freshwater lakes. Proc Natl Acad Sci USA Saving freshwater from salts. Science 351(6276):914–916 114(17):4453–4458 Chen CC, Zhu Y, Evans LB (1989) Phase Partitioning of Bio- Evans A, Zelazny LW, Zipper CE (1998) Solution parameters molecules: Solubilities of Amino Acids. Biotechnol Prog influencing dissolved organic carbon levels in three forest 5(3):111–118 soils. Soil Sci Soc Am J 52(6):1789–1792 Clavero V, Fernandez JA, Niell FX (1990) Influence of salinity Findlay SEG, Kelly VR (2011) Emerging indirect and long-term on the concentration and rate of interchange of dissolved road salt effects on ecosystems. Ann N Y Acad Sci phosphate between water and sediment in Fuente Piedra 1223:58–68 lagoon (S. Spain). Hydrobiologia 197(1):91–97 Fox LE (1991) The transport and composition of humic sub- Compton JE, Church MR (2011) Salt additions alter short-term stances in estuaries. In: Baker RA (ed) Human and Soils, nitrogen and carbon mobilization in a coastal oregon vol 1. Lewis Publishers, Chelsea MI, pp 129–162 andisol. J Environ Qual 40(5):1601–1606 Green SM, Machin R, Cresser MS (2008) Effect of long-term Conley DJ (2002) Terrestrial ecosystems and the global bio- changes in soil chemistry induced by road salt applications geochemical silica cycle. Global Biogeochem Cycles on N-transformations in roadside soils. Environ Pollut 16(4):68-1–68-8 152(1):20–31 Conway TM (2007) Impervious surface as an indicator of pH Green SM, Machin R, Cresser MS (2009) Does road slating and specific conductance in the urbanizing coastal zone of induce or ameliorate DOC mobilisation from roadside soils New Jersey, USA. J Environ Manag 85:308–316 to surface waters in the long term? Environ Monit Assess Cooper CA, Mayer PM, Faulkner BR (2014) Effects of road 153:435–448 salts on groundwater and surface water dynamics of Groffman PM, Law NL, Belt KT et al (2004) Nitrogen fluxes sodium and chloride in an urban restored stream. Biogeo- and retention in urban watershed Ecosystems. Ecosystems chemistry 121(1):149–166 7(4):393–403 Corsi SR, Graczyk DJ, Geis SW et al (2010) A fresh look at road Gutchess K, Jin L, Ledesma JLJ et al (2018) Long-term climatic salt: aquatic toxicity and water-quality impacts on local, and anthropogenic impacts on streamwater salinity in New regional, and national scales. Environ Sci Technol York State: INCA simulations offer cautious optimism. 44(19):7376–7382 Environ Sci Technol 52(3):1339–1347 Corsi SR, De Cicco LA, Lutz MA et al (2015) River chloride Hale RL, Groffman PM (2006) Chloride effects on nitrogen trends in snow-affected urban watersheds: increasing dynamics in forested and suburban stream debris dams. concentrations outpace urban growth rate and are common J Environ Qual 35(6):2425–2432 among all seasons. Sci Total Environ 508:488–497 Hedges JI, Keil RG (1999) Organic geochemical perspectives Cunningham MA, Snyder E, Yonkin D et al (2008) Accumu- on estuarine processes: sorption reactions and conse- lation of deicing salts in soils in an urban environment. quences. Mar Chem 65(1–2):55–65 Urban Ecosyst 11(1):17–31 Herbert ER, Boon P, Burgin AJ et al (2015) A global perspective Daniel MHB, Montebelo AA, Bernardes MC et al (2002) Effects on wetland salinization: ecological consequences of a of urban sewage on dissolved oxygen, dissolved inorganic growing threat to freshwater wetlands. Ecosphere and organic carbon, and electrical conductivity of small 6(10):1–43 streams along a gradient of urbanization in the Piracicaba Homer CG, Dewitz JA, Yang L et al (2015) Completion of the River Basin. Water Air Soil Pollut 136(1–4):189–206 2011 National Land Cover Database for the conterminous Devereux OH, Prestegaard KL, Needelman BA et al (2010) United States—representing a decade of land cover change Suspended-sediment sources in an urban watershed, information. Photogramm Eng Remote Sens Northeast Branch Anacostia River, Maryland. Hydrol 81(5):345–354 Process 24(11):1391–1403 Hupp CR, Noe GB, Schenk ER et al (2013) Recent and historic Dove PM, Elston SF (1992) Dissolution kinetics of quartz in sediment dynamics along difficult run, a suburban Virginia sodium chloride solutions: analysis of existing data and a Piedmont stream. Geomorphology 180–181:156–169 rate model for 25 C. Geochim Cosmochim Acta Jun M, Altor AE, Craft CB (2013) Effects on increased salinity 56(12):4147–4156 and inundation on inorganic nitrogen exchange and phos- Duan S, Kaushal SS (2015) Salinization alters fluxes of biore- phorus sorption by tidal freshwater floodplain forest soils, active elements from stream ecosystems across land use. Georgia (USA). Estuaries Coasts 36:508–518 Biogeosciences 12:7331–7347 Kaushal SS (2016) Increases salinization decreases safe drink- Duan Z, Sun R (2003) An improved model calculating CO2 ing water. Environ Sci Technol 50:2765–2766 solubility in pure water and aqueous NaCl solutions from

123 Biogeochemistry (2018) 141:463–486 485

Kaushal SS, Belt KT (2012) The urban watershed continuum: McClain ME, Boyer EW, Dent CL et al (2003) Biogeochemical evolving spatial and temporal dimensions. Urban Ecosys- hot spots and hot moments at the interface of terrestrial and tems 15(2):409–435 aquatic ecosystems. Ecosystems 6(4):301–312 Kaushal SS, Groffman PM, Likens GE et al (2005) Increased Merrikhpour H, Jalali M (2012) The effects of road salt appli- salinization of fresh water in the northeastern United cation on the accumulation and speciation of cations and States. Proc Natl Acad Sci USA 102(38):13517–13520 anions in an urban environment. Water Environ J Kaushal SS, Groffman PM, Band LE et al (2011) Tracking 27(4):524–534 nonpoint source nitrogen pollution in human-impacted Meybeck M (2003) Global occurrence of major elements in watersheds. Environ Sci Technol 45(19):8225–8232 rivers. In: Holland HD, Turekian KK (eds) Surface and Kaushal SS, Likens GE, Utz RM et al (2013) Increased River ground water, weathering, and soils. Treatise on geo- Alkalinization in the Eastern U.S. Environ Sci Technol chemistry, vol 5, 1st edn. Elsevier, Amsterdam, 47(18):10302–10311 pp 207–223 Kaushal SS, Mayer PM, Vidon PG et al (2014a) Land use and Miller CV, Chanat JG, Bell JM (2013) Water quality in the climate variability amplify carbon, nutrient, and contami- Anacostia River, Maryland and Rock Creek, Washington, nant pulses: a review with management implications. J Am D.C.: Continuous and discrete monitoring with simulations Water Resour Assoc 50(3):585–614 to estimate concentrations and yields of nutrients, sus- Kaushal SS, McDowell WH, Wollheim WM (2014b) Tracking pended sediment, and bacteria. Open-File Report evolution of urban biogeochemical cycles: past, present, 2013-1034. U.S. Geological Survey, Reston and future. Biogeochemistry 121(1):1–21. https://doi.org/ Newcomer-Johnson TA, Kaushal SS, Mayer PM et al (2014) 10.1007/s10533-014-0014-y Effects of stormwater management and stream restoration Kaushal SS, Duan S, Doody TR et al (2017) Human-accelerated on watershed nitrogen retention. Biogeochemistry weathering increases salinization, major ions, and alka- 121(1):81–106 linization in fresh water across land use. Appl Geochem Nieder R, Benbi DK, Scherer HW (2011) Fixation and defixa- 83:121–135 tion of ammonium in soils: a review. Biol Fertil Soils Kaushal SS, Likens GE, Pace ML et al (2018) Freshwater 47(1):1–14 salinization syndrome on a continental scale. Proc Natl Norrstrom AC (2005) Metal mobility by de-icing salt from an Acad Sci USA 115(4):E574–E583 infiltration trench for highway runoff. Appl Geochem Kelly VR, Lovett GM, Weathers KC et al (2008) Long-term 20(10):1907–1919 sodium chloride retention in a rural watershed: legacy Norrstrom AC, Bergstedt E (2001) The impact of road de-icing effects of road salt on streamwater concentration. Environ salts (NaCl) on colloid dispersion and base cation pools in Sci Technol 42(2):410–415 roadside soils. Water Air Soil Pollut 127(1–4):281–299 Kim S, Koretsky C (2011) Influence of NaCl and CaCl2 on lake Novotny V, Muehring D, Zitomer DH et al (1998) Cyanide and sediment biogeochemistry. Appl Geochem 26:S198–S201 metal pollution by urban snowmelt: Impact of deicing Kim S, Koretsky C (2013) Effects of road salt deicers on sedi- compounds. Water Sci Technol 38(10):223–230 ment biogeochemistry. Biogeochemistry Ondrasek G, Rengel Z, Domic R et al (2012) Salinity decreased 112(1–3):343–358 dissolved organic carbon in the rhizosphere and increases Kipton H, Powell J, Town RM (1992) Solubility and fraction- trace element phyto-accumulation. Eur J Soil Sci ation of humic acid—effect of pH and ionic medium. Anal 63(5):685–693 Chim Acta 267(1):47–54 Oren A (2001) The bioenergetic basis for the decrease in Li X, Rengel Z, Mapfumo E et al (2007) Increase in pH stim- metabolic diversity at increasing salt concentrations: ulates mineralization of ‘native’ organic carbon and implications for the functioning of salt lake ecosystems. nitrogen in naturally salt-affected sandy soils. Plant Soil Hydrobiologia 466(1–3):61–72 290(1–2):269–282 Pellerin BA, Saraceno JF, Shanley JB et al (2012) Taking the Li M, Zhao L, Zhang J (2013) Effect of temperature, pH, and salt pulse of snowmelt: in situ sensors reveal seasonal, event on fluorescent quality of water extractable organic matter and diurnal patterns of nitrate and dissolved organic matter in black soil. J Integr Agric 12(7):1251–1257 variability in an upland forest stream. Biogeochemistry Lofgren S (2001) The chemical effects of deicing salt on soil and 108(1–3):183–198 stream water of five catchments in Southeast Sweden. Pellerin BA, Bergamaschi BA, Downing BD et al (2013) Water Air Soil Pollut 130(1–4):863–868 Optical techniques for the determination of nitrate in Majcher EH, Woytowitz EL, Reisinger AJ et al (2018) Factors environmental waters: guidelines for instrument selection, affecting long-term trends in surface-water quality in the operation, deployment, maintenance, quality assurance, Gwynns Falls watershed, Baltimore City and County, and data reporting. U.S. Geological Survey Techniques and Maryland, 1998-2016. Open-File Report 2018-1038. U.S. Methods 1-D5 Geological Survey, Reston Pennino MJ, Kaushal SS, Mayer PM et al (2016) Stream Marsalek J (2003) Road salts in urban stormwater: an emerging restoration and sewers impact sources and fluxes of water, issue in stormwater management in cold climates. Water carbon, and nutrients in urban watersheds. Hydrol Earth Sci Technol 48(9):61–70 Syst Sci 20:3419–3439 Mayer PM, Groffman PM, Striz EA et al (2010) Nitrogen Price JR, Szymanski DW (2014) The effects of road salt on dynamics at the groundwater-surface water interface of a stream water chemistry in two small forested watersheds, degraded urban stream. J Environ Qual 39(3):810–823 Catoctin mountain, Maryland, USA. Aquat Geochem 20(2–3):243–265 123 486 Biogeochemistry (2018) 141:463–486

Radtke DB, Davis JV, Widle FD (2005) 6.3 Specific electrical oxyhydroxides along pH and salinity gradients in estuaries conductance. US Geological Survey National Field Man- and coastal aquifers. Geochim Cosmochim Acta ual for the Collection of Water-quality Data, TWRI book 9 72(14):3431–3445 Ramakrishna DM, Viraraghavan T (2005) Environmental Srividya S, Soumya S, Pooja K (2009) Influence of environ- impact of chemical deicers—a review. Water Air Soil mental factors and salinity on phosphate solubilization by a Pollut 166(1–4):49–63 newly isolated Aspergillus niger F7 from agricultural soil. Robinson HK, Hasenmueller EA, Chambers LG (2017) Soil as a Afr J Biotech 8(9):1864–1870 reservoir for road salt retention leading to its gradual Steele MK, Aitkenhead-Peterson JA (2011) Long-term sodium release to groundwater. Appl Geochem 83:72–85 and chloride surface water exports from the Dallas/Fort Rode M, Wade AJ, Cohen MJ et al (2016) Sensors in the stream: Worth region. Sci Total Environ 409(16):3021–3032 the high-frequency wave of the present. Environ Sci Steele MK, Aitkenhead-Peterson JA (2012) Urban soils of Technol 50(19):10297–10307 texas: relating irrigation sodicity to water-extractable car- Romero-Aranda MR, Jurado O, Cuartero J (2006) Silicon alle- bon and nutrients. Soil Sci Soc Am J 76(3):972–982 viates the deleterious salt effect on tomato plant growth by Steele MK, Aitkenhead-Peterson JA (2013) Salt impacts on improving plant water status. J Plant Physiol organic carbon and nitrogen leaching from senesced veg- 163(8):847–855 etation. Biogeochemistry 112(1–3):245–259 Saiers JE, Lenhart JJ (2003) Ionic-strength effects on colloid Stets EG, Lee CJ, Lytle DA et al (2018) Increasing chloride in transport and interfacial reactions in partially saturated rivers of the conterminous U.S. and linkages to potential porous media. Water Resour Res 39(9):1256 corrosivity and lead action level exceedances in drinking Shanley JB (1994) Effects of ion exchange on stream solute water. Sci Total Environ 613–614:1498–1509 fluxes in a basin receiving highway deicing salts. J Environ Tanford C (1961) Physical chemistry of macromolecules. Qual 23(5):977–986 Wiley, New York Shi Y, Wang Y, Flowers TJ (2013) Silicon decreases chloride Treguer P, Nelson DM, Van Bennekom AJ et al (1995) The transport in rice (Oryza sativa L.) in saline conditions. silica balance in the world ocean: a reestimate. Science J Plant Physiol 170(9):847–853 268(5209):375–379 Shields CA, Band LE, Law N et al (2008) Streamflow distri- Weiss RF (1974) Carbon dioxide in water and seawater: the bution of non-point source nitrogen export from urban- solubility of a non-ideal gas. Mar Chem 2(3):203–215 rural catchments in the Chesapeake Bay watershed. Water Weston NB, Dixon RE, Joye SB (2006) Ramifications of Resour Res 44(9):W09416. https://doi.org/10.1029/ increased salinity in tidal freshwater sediments: geo- 2007WR006360 chemistry and microbial pathways of organic matter min- Sholkovitz ER (1976) Flocculation of dissolved organic and eralization. J Geophys Res Biogeosci 111(G1):G01009. inorganic matter during the mixing of river water and https://doi.org/10.1029/2005JG000071 seawater. Geochim Cosmochim Acta 40(7):831–845 Williams WD (2001) Anthropogenic salinisation of inland Smith RM, Kaushal SS (2015) Carbon cycle of an urban waters. Hydrobiologia 466:329–337 watershed: exports, sources, and metabolism. Biogeo- Woo TJ, Salice CJ (2017) Timing is everything: pulsed versus chemistry 126(1–2):173–195 constant exposures in assessing effects of road salt on Snodgrass JW, Moore J, Lev SM et al (2017) Influence of aquatic organisms. Integr Environ Assess Manag modern stormwater management practices on transport of 13(4):792–794 road salt to surface waters. Environ Sci Technol Zhang J, Huang X (2011) Effects of temperature and salinity on 51(8):4165–4172 phosphate sorption on marine sediments. Environ Sci Spiteri C, Van Cappellen P, Regnier P (2008) Surface com- Technol 45(16):6831–6837 plexation effects on phosphate adsorption to ferric iron

123